How are you bringing up the gbe interface? Run netdiag and see if you
get an error like the below which I got from one of our filers:
filer> netdiag
Performing physical layer diagnostics.....
You may have a duplex mismatch on the link e0
Configured mediatype: auto-100tx-fd-up
Corresponding switch port may be configured half duplex/auto
.
.
.
I believe the recommended way to bring up a gbe interface is the below
for the controller version I, II, and III (as seen in sysconfig -v):
ifconfig e6 `hostname` mediatype auto flowcontrol full netmask 255.255.255.0
I don't know about controller version IV or later but you may want to
try a mediatype of auto-1000fx or 1000fx instead of auto. NetApp has
always recommended to hard set the speeds and flow for both the switch
port and the filer for 10/100 connections, but it looks like they
want to auto-neg in gbe land.
You may also want to check the TCP window sizes...or does that just
affect NFS...I should really know that.
Jeff
On Tue, Jun 24, 2003 at 11:43:37AM -0400, Holland, William L wrote:
I'm doing some performance testing with an F810 that is connected to our
network via the onboard 10/100 NIC. DNS, WINS, etc are configured on that
interface. The F810 also has a dual port GbE NIC connected directly to a
Win2K server that is also connected via 10/100 to the same network as the
F810's onboard 10/100. When I try things such as
rsh 192.168.2.1 date
it takes a VERY long time to run. Also connecting via unc (e.g.
\\192.168.2.1\c$) takes a lot longer than it does over the 10/100 network.
I suspect the issue is really on the Win2K side and how it tries to find
clients but I am not sure what to look at.
Any ideas?
Thanks in advance.